ecDNA-driven oncogene super-expressors shape immunoevasive tumor microenvironment

ecDNA-driven oncogene super-expressors shape immunoevasive tumor microenvironment | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results ecDNA-driven oncogene super-expressors shape immunoevasive tumor microenvironment View ORCID Profile Kailiang Qiao , View ORCID Profile Qing-Lin Yang , View ORCID Profile Tuo Li , View ORCID Profile Xiongfeng Chen , View ORCID Profile Zeynep Yazgan , View ORCID Profile Yoon Jung Kim , View ORCID Profile Collin Gilbreath , View ORCID Profile Jun Yi Stanley Lim , View ORCID Profile Yipeng Xie , View ORCID Profile Xiaohui Sun , View ORCID Profile Yang Liu , View ORCID Profile Yiyue Jia , View ORCID Profile Zhijian J. Chen , View ORCID Profile Huocong Huang , View ORCID Profile Sihan Wu doi: https://doi.org/10.1101/2025.11.15.688565 Kailiang Qiao 1 Children’s Medical Center Research Institute, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Kailiang Qiao Qing-Lin Yang 1 Children’s Medical Center Research Institute, University of Texas Southwestern Medical Center , Dallas, TX, USA 2 Department of Molecular Biology, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Qing-Lin Yang Tuo Li 2 Department of Molecular Biology, University of Texas Southwestern Medical Center , Dallas, TX, USA 3 Center for Inflammation Research, University of Texas Southwestern Medical Center , Dallas, TX, USA 4 Howard Hughes Medical Institute, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Tuo Li Xiongfeng Chen 5 Department of Surgery, University of Texas Southwestern Medical Center , Dallas, TX, USA 6 Hamon Center for Therapeutic Oncology Research, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Xiongfeng Chen Zeynep Yazgan 5 Department of Surgery, University of Texas Southwestern Medical Center , Dallas, TX, USA 6 Hamon Center for Therapeutic Oncology Research, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Zeynep Yazgan Yoon Jung Kim 1 Children’s Medical Center Research Institute, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Yoon Jung Kim Collin Gilbreath 1 Children’s Medical Center Research Institute, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Collin Gilbreath Jun Yi Stanley Lim 1 Children’s Medical Center Research Institute, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Jun Yi Stanley Lim Yipeng Xie 1 Children’s Medical Center Research Institute, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Yipeng Xie Xiaohui Sun 1 Children’s Medical Center Research Institute, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Xiaohui Sun Yang Liu 7 Department of Immunology, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Yang Liu Yiyue Jia 5 Department of Surgery, University of Texas Southwestern Medical Center , Dallas, TX, USA 6 Hamon Center for Therapeutic Oncology Research, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Yiyue Jia Zhijian J. Chen 2 Department of Molecular Biology, University of Texas Southwestern Medical Center , Dallas, TX, USA 3 Center for Inflammation Research, University of Texas Southwestern Medical Center , Dallas, TX, USA 4 Howard Hughes Medical Institute, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Zhijian J. Chen For correspondence: zhijian.chen{at}utsouthwestern.edu huocong.huang{at}utsouthwestern.edu sihan.wu{at}utsouthwestern.edu Huocong Huang 5 Department of Surgery, University of Texas Southwestern Medical Center , Dallas, TX, USA 6 Hamon Center for Therapeutic Oncology Research, University of Texas Southwestern Medical Center , Dallas, TX, USA 7 Department of Immunology, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Huocong Huang For correspondence: zhijian.chen{at}utsouthwestern.edu huocong.huang{at}utsouthwestern.edu sihan.wu{at}utsouthwestern.edu Sihan Wu 1 Children’s Medical Center Research Institute, University of Texas Southwestern Medical Center , Dallas, TX, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Sihan Wu For correspondence: zhijian.chen{at}utsouthwestern.edu huocong.huang{at}utsouthwestern.edu sihan.wu{at}utsouthwestern.edu Abstract Full Text Info/History Metrics Supplementary material Preview PDF Abstract ecDNA contributes to cancer genetic heterogeneity through random segregation during mitosis. Emerging evidence links ecDNA to immune evasion, but the mechanism remains elusive. Using genetically engineered mouse models of pancreatic ductal adenocarcinoma (PDAC), we show that Kras and Myc oncogenes are amplified either on ecDNAs or as homogeneously staining regions (HSRs) on chromosomes. ecDNA-driven tumors are more aggressive in immunocompetent mice. Single-cell transcriptomic and histological analyses reveal that ecDNA-driven tumors rapidly establish an immunoevasive tumor microenvironment (TME), marked by increased myofibroblastic cancer-associated fibroblasts (myCAFs) and reduced T cell infiltration. Mechanistically, ecDNA heterogeneity generates a subset of cancer cells with extremely high Kras expression, termed super-expressors, which secrete amphiregulin to promote myCAF expansion and suppress T cell infiltration. Clonally organized super-expressors establish an immunoevasive niche in the TME from patients with PDAC. Our findings demonstrate a causal role of ecDNA in TME remodeling, offering insights into cancer heterogeneity and immune evasion. Introduction Cancers possess multiple mechanisms to evade immune surveillance and destruction. Cancer cell-intrinsic alterations, such as genetic, epigenetic, and metabolic aberrations, profoundly influence the composition and function of the tumor microenvironment (TME). 1 , 2 For example, many canonical oncogenes, such as KRAS and MYC , have been shown to facilitate immune evasion by regulating the expression of cytokines, chemokines, and immune checkpoint molecules. 3 – 7 Consequently, the cell composition and cell-cell communication in the TME are reshaped to be more immunosuppressive, characterized by decreased T cell infiltration, increased expansion of cancer-associated fibroblasts (CAFs), and, eventually, reduced efficacy of immune therapies. Extrachromosomal DNA (ecDNA) represents an emerging hallmark of aggressive cancer. Its non-Mendelian inheritance, high copy number, and hyper-accessible chromatin drive profound oncogene amplification and intratumoral heterogeneity. 8 , 9 Patients with ecDNA-driven cancers usually have poor clinical outcomes, including therapy resistance and shorter survival. 10 , 11 More recently, ecDNA amplification has been increasingly associated with immunoevasive TME. Two pioneering pan-cancer analyses using The Cancer Genome Atlas (TCGA) data, based on bulk-cell RNA sequencing (RNA-seq), independently demonstrate that ecDNA-containing tumors exhibit an immune-cold phenotype, characterized by reduced immune cell infiltration, downregulated immunoregulatory pathways, and diminished antigen presentation. 12 , 13 Aligned with these findings, a single-cell RNA sequencing (scRNAseq) study reveals that urothelial carcinomas harboring ecDNA display decreased MHC-I expression in malignant cells and increased regulatory T cell (Treg) infiltration compared to tumors lacking focal somatic copy number amplifications. 14 A spatial transcriptomic study in pancreatic ductal adenocarcinoma (PDAC) further links high MYC expression, likely through the amplification of ecDNA, to the enrichment of CD90+ myofibroblastic CAFs (myCAFs) and the depletion of cytotoxic T cells. 15 Notably, ecDNA may also harbor immunomodulatory genes independent of canonical oncogenes, suggestive of positive selection of immunoevasive ecDNA species. 16 – 18 Bulk-cell transcriptomic analysis shows that tumors carrying such immunomodulatory ecDNAs exhibit significantly reduced T cell infiltration compared to those harboring oncogene-only ecDNAs. 18 While accumulating evidence links ecDNA to an immunoevasive phenotype, the causality underlying this association remains elusive: Does ecDNA shape an immunosuppressive TME, or does a more immunosuppressive TME favor ecDNA pathogenesis? Addressing this question faces significant challenges. First, clinical data relying on bulk-cell sequencing cannot accurately deconvolute the cellular complexity of the TME. Second, while emerging single-cell and spatial genomic technologies offer high-resolution insights, establishing causality in human tissues remains a formidable challenge. This difficulty stems from the lack of isogenic controls, which makes it extremely hard to disentangle the specific effects of ecDNA from other forms of chromosomal amplification. Finally, while mouse ecDNA cancer models become available, 19 current studies have yet to generate isogenic controls comparing ecDNA to chromosomal amplification. To address these limitations, we used the well-established KPfC mouse PDAC model ( Kras LSL-G12D/+ ; Trp53 fl/fl ; Pdx1 Cre/+ ) to generate primary isogenic cancer cell clones. 20 – 22 Surprisingly, we found that Kras and Myc oncogenes were spontaneously amplified as ecDNAs and homogeneously staining regions (HSRs) on chromosomes. From clones typically harboring both forms of amplification, we isolated isogenic pairs carrying exclusively ecDNA or HSR. Using scRNAseq, genetic knockout, and histological analyses for tumors in immunocompetent syngeneic mice, together with spatial transcriptomic analyses in PDAC patient samples, we investigated the causal role of ecDNA amplification in shaping an immunoevasive TME. Results KPfC PDAC spontaneously acquires ecDNA and HSR amplification To establish a comprehensive library of congenic cancer cell clones, we harvested fully developed tumors that had been raised spontaneously in the KPfC mouse and processed them into single-cell suspensions through enzymatic digestion. Individual clones were isolated through limiting dilution and subsequently expanded through serial passaging. Each established clone was rigorously characterized and validated for tumorigenic potential through subcutaneous injection into immunocompetent syngeneic recipients, confirming their ability to form tumors in vivo . This library of clones recapitulates the heterogeneity of cancer cell populations present within the entire tumor, capturing distinct lineage phenotypes that represent the diverse cellular subpopulations found in the spontaneous tumors. This approach yielded a genetically homogeneous yet phenotypically diverse collection of tumor cell clones that retained the key molecular characteristics of the parental tumors ( Figure 1A ). Download figure Open in new tab Figure 1. Establishment and characterization of ecDNA and HSR-containing PDAC cell lines. (A) Flowchart depicting the process to isolate and identify ecDNA or HSR-containing isogenic PDAC cell clones from tumors in KPfC mice. (B) Kras and Myc DNA FISH at metaphase spreads in EC1/2 and HSR1/2 in vitro cultured cells from the primary tumor dissociation. Scale bar: 5 μm. (C) Workflow for validating the maintenance of ecDNA and HSR amplicons in orthotopic tumors derived from EC1/2 and HSR1/2 clones. (D) Kras and Myc DNA FISH at metaphase spreads in EC1/2 and HSR1/2 in vitro cultured cells from the secondary tumor dissociation. Scale bar: 5 μm. (E) Subcutaneous tumor growth curve of EC1/2 and HSR1/2 isogenic clones. Two-way ANOVA with Tukey’s HSD. NS indicates not significant. Sample size: EC1, n = 9; EC2, n = 8; HSR1, n = 9; HSR2, n = 10. (F) Overall survival of C57BL/6J mice bearing EC1/2 or HSR1/2 tumors. Pairwise log-rank test with false discovery rate correction. Sample size: EC1, n = 8; EC2, n = 8; HSR1, n = 10; HSR2, n = 10. We identified one clone, designated CT1BA5, that harbored both Kras ecDNAs and HSRs. 23 We then isolated two ecDNA-only and two HSR-only isogenic subclones from the parental CT1BA5 cells, termed EC1 or EC2 and HSR1 or HSR2. All isogenic subclones were verified by Kras DNA fluorescence in situ hybridization (FISH) on metaphase chromosome spreads ( Figure 1B ). Whole-genome sequencing coupled with AmpliconArchitect analysis showed that these isogenic clones bore nearly identical Kras amplicon structures (Figure S1). Allele frequency analysis showed that the majority of Kras amplicons harbored the G12D mutation, suggesting a positive selection for the mutant Kras allele (Figure S2A). AmpliconArchitect additionally revealed that the Myc locus was also amplified on another ecDNA species, which was subsequently verified by Myc DNA FISH, showing Myc ecDNA or HSR amplification, respectively, in the EC1/2 or HSR1/2 clones ( Figure 1B , Figure S1). The Kras and Myc amplification status of these lines was stably maintained in orthotopic tumors, as confirmed by DNA FISH on metaphase spreads ( Figures 1C-D ). To investigate the transcriptional profile of these clones, we performed RNA-seq and found that EC1/2 and HSR1/2 exhibited similar phenotypes in vitro . Analyses of principal components and differentially expressed genes based on RNA sequencing data revealed similar transcriptional patterns among individual clones within the EC or HSR group, with only 1% of the transcriptome contributing to the difference between EC and HSR (Figures S2B-C). Gene set enrichment analysis (GSEA) demonstrated an insignificant difference in the KRAS signaling, and only one of the MYC signaling signatures was marginally upregulated in EC clones (Figure S2D). Moreover, all four clones showed negligible differences in cell proliferation capacity in vitro (Figure S2E). These data suggest that the variation of Kras and Myc copy number among clones at the bulk-cell level does not drive a significant phenotypic contrast in vitro . However, tumors derived from EC1/2 cells grew more rapidly than those from HSR1/2 cells in immunocompetent C57BL/6J mice ( Figure 1E ). Consequently, mice bearing EC tumors exhibited significantly shorter survival times ( Figure 1F ). These results show that ecDNA-containing cancer cells promote more aggressive tumor progression than their isogenic HSR counterparts, likely through interactions with the TME rather than solely cell-intrinsic mechanisms. ecDNA-driven PDAC rapidly establishes an immunoevasive TME Next, we investigated TME differences between EC and HSR tumors. We established orthotopic tumors using EC1 or HSR1 cells and harvested them at early- and mid-stages (10- and 15-days post-inoculation, respectively) ( Figure 2A ). To characterize the TME, we performed scRNAseq on mid-stage tumors from both groups. At this time point, tumor weights showed no significant difference (Figure S3A), thereby avoiding artifacts caused by size disparities or necrosis associated with larger tumor volumes. To minimize sampling bias, two bisected tumors within each group were randomly combined. A total of 3 samples for each group were dissociated to prepare single-cell suspensions ( Figure 2A ). Following enzymatic digestion, we captured approximately 20,000 cells per tumor type. Reference-based annotation using SingleR identified the cell composition of the TME, 24 including cancer cells, immune cells, fibroblasts, and endothelial cells ( Figure 2B ), whose identities were verified using canonical biomarkers (Figure S3B). Download figure Open in new tab Figure 2. scRNAseq profiling of ecDNA and HSR-driven PDAC TME with IHC validation. (A) Schematic overview of the scRNAseq workflow. (B) UMAP visualization of all cells identified by scRNAseq from EC1 and HSR1 tumors. (C) Cell composition analysis of stromal and immune cell populations. Fisher’s exact test with FDR correction. NS indicates not significant. (D) UMAP visualization of CAF subclusters. (E) Cell composition analysis of CAF subclusters. Colors correspond to those in Figure 2D . Fisher’s exact test for the overall p-value, pairwise Fisher’s exact test with Benjamini-Hochberg correction for specific cell types. (F) Representative images of POSTN IHC staining showing both low- and high-magnification views. Scale bars 2.5 mm (low magnification), 50 μm (high magnification). (G) Quantification of POSTN IHC staining. Each color within a group represents an individual mouse. Four non-overlapping regions from the same section were analyzed and shown in the same color. Sample sizes: EC, n = 6; HSR, n = 6. Wilcoxon test. (H) Cell composition analysis of all Ptprc + immune cells. Fisher’s exact test with FDR correction. NS indicates not significant. (I) Representative images of CD3E IHC staining showing low- and high-magnification views. Scale bars: 5 mm (low magnification) and 50 μm (high magnification). Red letter T and circles indicate the tumor area in the pancreas. (J) Quantification of CD3E IHC staining. Each color within a group represents an individual mouse. Four non-overlapping regions from the same section were analyzed and shown in the same color. Sample sizes: EC, n = 7; HSR, n = 8. Wilcoxon test. (K) Representative images of CD8A IHC staining showing low- and high-magnification views. Scale bars: 5 mm (low magnification) and 50 μm (high magnification). Red letter T and circles indicate the tumor area in the pancreas. (L) Quantification of CD8A IHC staining. Each color within a group represents an individual mouse. Four non-overlapping regions from the same section were analyzed and shown in the same color. Sample sizes: EC, n = 7; HSR, n = 8. Wilcoxon test. Focusing on stromal and immune cell populations, EC tumors contained significantly more fibroblasts and slightly fewer immune cells compared to HSR tumors ( Figure 2C ). To characterize fibroblast heterogeneity, we extracted and re-clustered all fibroblasts from the dataset. Using established cancer-associated fibroblasts (CAFs) subtype markers, we identified inflammatory CAFs (iCAFs), myofibroblastic CAFs (myCAFs), antigen-presenting CAFs (apCAFs), proliferating CAFs (a proliferative myCAF subset), and mesothelial cells ( Figures 2D , Figure S3C). 25 – 28 EC tumors exhibited increased proportions of myCAFs and proliferating CAFs but decreased apCAFs ( Figure 2E ). As CAF networks are particularly resistant to enzymatic digestion due to dense extracellular matrix, our scRNAseq may underestimate CAF abundance. We therefore validated these findings using immunohistochemistry (IHC) for periostin (POSTN), a myCAF marker, confirming significantly higher myCAF composition in EC tumors ( Figures 2F-G ). Within the immune cell populations, EC tumors contained fewer T and B cells ( Figure 2H ). We extracted and re-clustered all T cells, identifying CD8+ T cells and Tregs based on Cd4 , Foxp3 , and Ctla4 expression (Figures S3D-E). 29 EC tumors exhibited reduced CD8+ T cells but increased Tregs, indicating a more immunosuppressive TME (Figure S3F). However, total T cell abundance was scarce in both EC and HSR tumors (<2.6% of TME cells), indicating that both groups exhibited an immune-cold phenotype by mid-stage (Figure S3G). To assess the onset of immunoevasion, we examined early-stage tumors (10 days post-inoculation) by IHC for CD3E and CD8A. HSR tumors showed abundant T cell infiltration, while EC tumors already exhibited significantly reduced T cell infiltration ( Figures 2I-L ). These findings indicate that EC tumors establish an immunoevasive TME rapidly after tumor initiation, preceding the immune-cold transition observed in HSR tumors. ecDNA heterogeneity drives Kras super-expressors Next, we investigated the mechanism by which ecDNA-driven PDAC shapes an immunoevasive TME through myCAF expansion and T cell suppression. In contrast to focal gene amplification on chromosomes, ecDNAs drive oncogene copy number heterogeneity through unequal mitotic segregation, thereby generating a pool of genetically diverse cancer cells for selection. 30 While cancer heterogeneity has been linked to aggressive phenotypes, the mechanisms by which ecDNAs foster a more immunoevasive TME remain unclear. We hypothesized that a small subset of ecDNA-high or ecDNA-low cell populations may have a distinct transcriptional profile, which altered communication between cancer cells and host cells. Therefore, we profiled the Kras expression level for EC and HSR tumors. Among cancer cells, the lower bound of the Kras expression was comparable between EC and HSR tumors; however, EC cancer cells exhibited a markedly higher upper bound and greater variation of Kras expression ( Figure 3A ), as assessed by the median absolute deviation (MAD) and interquartile range (IQR). This pattern was consistently observed at the levels of DNA copy number and protein abundance, with Kras DNA FISH and protein immunofluorescence (IF) signals showing elevated upper bounds and greater variation in EC tumors (Figures S4A-D). Co-staining of Kras DNA FISH and RAS protein IF revealed a positive correlation between Kras ecDNA copy number and protein abundance (Figures S4E-F), indicating that expression variation of Kras was dictated by ecDNA copy number heterogeneity. Download figure Open in new tab Figure 3. Molecular characteristics of ecDNA-driven Kras super-expressors. (A) Density plot of Kras mRNA expression of EC and HSR cancer cells from scRNAseq data. The distribution variance was assessed using median absolute deviation (MAD) and interquartile range (IQR). The vertical dashed lines indicate the 95% percentile, which was used to distinguish super- and normal-expressors. Number of cells analyzed: EC, n = 10,000; HSR, n = 10,000. (B) Proportion analysis of normal-expressor and super-expressor in EC and HSR groups based on the same subset of cells in Figure 3A . (C) Differential pathway analysis using SCPA with the mouse-ortholog hallmark gene set as input. (D) Differential pathway analysis using GSEA with the stress-related pathways from mouse collections as input. (E) Schematic showing the input of GSEA analysis. The DEGs between EC and HSR transcriptomes were used to generate two gene sets: the Upregulated gene set in EC and the Downregulated gene set in EC. A ranked list of genes was generated based on expression differences between super- and normal-expressors. (F-G) GSEA results show the contrast between Kras super- and normal-expressors resembles the differences between EC and HSR cancer cells. (H) Correlation between the expression of the human orthologs of Kras super-expressor signature genes and the immune cell infiltration scores in the TCGA pancreatic cancer (TCGA-PAAD) cohort from TIMEDB, analyzed using the Pearson correlation test with FDR correction. In contrast, the variance of Myc expression did not show a substantial difference between EC and HSR cancer cells (Figure S4G). Furthermore, DNA FISH-IF analysis revealed a weaker correlation between Myc ecDNA copy number and protein abundance (Figures S4H-I). Notably, cancer cells with the highest MYC protein expression did not possess the highest Myc ecDNA copy number, suggesting that additional regulatory mechanisms beyond DNA copy number contribute to Myc expression in this cancer model. Taken together, these data suggest that Kras ecDNA may give rise to a subpopulation of cancer cells with exceptionally high Kras expression, hereinafter referred to as Kras super-expressors. Kras super-expressors are associated with an immunoevasive TME To investigate the functional role of ecDNA-driven Kras super-expressors, we performed pathway analysis using scRNAseq data (Figure S5A). An equal number of cancer cells (n = 10,000) was randomly sampled and pooled from EC and HSR tumors. Super- and normal-expressors were defined based on above or below the top 5% Kras expression level cutoff ( Figure 3A , Figure S5B). Under this threshold, KRAS signaling was significantly upregulated in super-expressors compared to normal-expressors within EC tumors, whereas normal-expressors between EC and HSR tumors showed no significant difference in KRAS signaling (Figure S5B). This strategy optimized discovery sensitivity while minimizing false positives. Under this definition, the vast majority (99%) of Kras super-expressors in the 1:1 mixed single-cell pool originated from EC tumors ( Figure 3B , Figure S5C). Single-cell pathway analysis using Single cell pathway analysis (SCPA) revealed that Kras super-expressors downregulated numerous immune response pathways, 31 including the interferon alpha and gamma pathways, as well as the TNF-alpha signaling ( Figure 3C ). However, these Kras super-expressors also experienced cellular stress, including replication stress and ER stress, as shown by GSEA ( Figure 3D ). For example, genes involved in single-stranded DNA binding proteins ( Rpa1 / 2 ), the 9-1-1 complex ( Rad9a / Rad1 /Hus1), and the check1-claspin axis ( Chek1 / Clspn ) were upregulated, indicating elevated replication stress (Figure S5D). Further analysis using Kras and Myc DNA FISH in EC tumors revealed that high copy numbers of Kras and Myc rarely coexisted within the same cancer cell (Figure S5E), suggesting a potential mutual exclusivity in ecDNA-driven amplification shaped by oncogene overdose. 32 These findings suggest that, while Kras super-expressors may evade immune surveillance better, elevated cellular stress may limit their proliferation (Figure S5F). To assess whether Kras super-expressors contribute to the phenotypic divergence between EC and HSR tumors, we performed GSEA to determine whether gene sets derived from differentially expressed genes between EC and HSR significantly changed between super- and normal-expressors transcriptomes ( Figure 3E ). The results revealed significant and concordant differences, supporting the hypothesis that ecDNA-driven Kras super-expressors shape the biological distinction between EC and HSR tumors ( Figures 3F-G ). To further explore the impact of Kras super-expressors on the TME, we analyzed the correlation between the expression of Kras super-expressor signature genes and the immune cell score in the TCGA pancreatic cancer (TCGA-PAAD) cohort from TIMEDB. 33 The top 30 signature genes (Figure S5A, Supplementary Table 1) were mapped to their human orthologs to obtain transcripts per million (TPM) values, and the geometric mean was calculated per subject. Correlation analysis revealed that the Kras super-expressor signature was associated with increased infiltration of CAFs, natural Tregs, and neutrophils, alongside decreased infiltration of CD8+ T cells and activated NK cells, mirroring the immunoevasive phenotype in EC tumors ( Figure 3H ). These data suggest that Kras super-expressors may play a pivotal role in establishing an immunoevasive TME. Kras super-expressors induce myCAF expansion via AREG signaling To identify effector genes in Kras super-expressors that suppressed the immune response, we applied a multi-step filtering strategy. Candidate genes were selected based on three criteria: (1) upregulated in Kras super-expressors compared to normal-expressors; (2) upregulated in TCGA-PAAD tumors relative to normal pancreatic tissues from the Genotype-Tissue Expression (GTEx) dataset; (3) annotated as cell-cell communication ligands from CellChat ( Figure 4A , Figure S5A). 34 This approach yielded five candidate genes, including AREG , CEACAM1 , CSF2 , PDCD1LG2 , and PTHLH ( Figures 4A-B , Figures S6A-B). Among these, upregulation of AREG and CSF2 was significantly associated with a shorter overall survival in the TCGA-PAAD cohort ( Figure 4C , Figure S6B). Download figure Open in new tab Figure 4. Functional validation of Areg in PDAC TME remodeling. (A) Venn diagram showing the identification of effector genes in Kras super-expressors. Upregulated genes between TCGA-PAAD tumors and GTEx normal pancreas samples were intersected with ligand genes from CellChat and Kras super-expressor upregulated genes from scRNAseq analysis. (B) Dot plot showing the differential expression of AREG between TCGA-PAAD tumor samples and GTEx normal samples. Wilcoxon test. (C) Overall survival of TCGA-PAAD cohort stratified by AREG expression (median 50% cutoff). Log-rank test. (D) Gross image of orthotopic tumors from C57BL/6J mice inoculated with CT1BA5EC1 cell lines with or without Areg knockout. NTC indicates non-targeting control. (E) Box plot showing the tumor weight difference among the three groups. Sample sizes: sgNTC, n = 7; sg Areg #1, n = 7; sg Areg #2, n = 7. One-way ANOVA with Tukey’s HSD. (F) Overall survival of C57BL/6J mice bearing CT1BA5EC1 tumors with or without loss of Areg . Sample sizes: sgNTC, n = 10; sg Areg #1, n = 9; sg Areg #2, n = 10. Pairwise log-rank test with FDR correction. NS indicates not significant. (G) Chord diagram generated by CellChat based on scRNAseq data showing the Areg-Egfr ligand-receptor communication among different cell types in EC samples. Each chord represents communication between two cell types. The direction of the chords is from the sender to the receiver. The width of a chord represents interaction strength. (H-M) Representative images of POSTN (H), CD3E (J), and CD8A (L) IHC staining with both low- and high-magnification views. Scale bars 5 mm (low magnification), 50 μm (high magnification). Dot plots show the quantification of POSTN (I), CD3E (K), and CD8A (M). Each color within a group represents an individual mouse. Four non-overlapping regions from the same section were analyzed and shown in the same color. Sample sizes: sgNTC, n = 7; sg Areg #1, n = 7; sg Areg #2, n = 7. Kruskal–Wallis test with Bonferroni correction. Csf2 (granulocyte-macrophage colony-stimulating factor, GM-CSF) is known to orchestrate an immunosuppressive PDAC TME by regulating myeloid-derived suppressor cells. 6 , 7 Areg encodes amphiregulin, an EGF-like growth factor that has been implicated in promoting metastasis via autocrine from myCAFs in PDAC. 35 We decided to focus on Areg because of its higher expression in our mouse models (Figures S6C-E). To confirm that KRAS signaling regulates Areg expression, we performed RNAi-mediated Kras knockdown in EC1 cells, which significantly reduced Areg expression (Figure S6F). Consistently, pharmacological inhibition of the MEK-ERK pathway using trametinib (MEKi) and SCH772984 (ERKi) dramatically suppressed Areg expression (Figures S6G-J). Inhibition of the mTOR signaling by rapamycin (mTORC1i) and torin 2 (mTORC1/2i) yielded a significant yet milder suppression effect on Areg expression (Figures S6G-J). To investigate the functional role of Areg in PDAC, we knocked out Areg in EC1 cells (Figure S7A). Areg deletion did not alter Kras or Myc ecDNA copy numbers, nor did it impair cell viability in vitro (Figures S7B-E). However, orthotopic implantation of Areg -deficient EC1 cells resulted in significantly reduced tumor growth and prolonged survival in mice ( Figures 4D-F ), suggesting that amphiregulin promotes PDAC malignancy through interactions with the TME rather than a cancer cell-intrinsic mechanism. scRNAseq analysis showed that Kras super-expressors exhibited the highest transcription of Areg , whereas the amphiregulin receptor Egfr was primarily expressed in fibroblasts (Figure S7F). Cell-cell communication analysis further revealed a stronger ligand-receptor interaction between Kras super-expressors (sender) and fibroblasts (receiver) in EC tumors, even though super-expressors accounted for only 5% of the cancer cells ( Figure 4G ). These findings prompted us to investigate the role of Areg in myCAF expansion. Knockout of Areg in EC1 cancer cells significantly reduced the number of myCAFs in EC tumors ( Figures 4H-I ). Moreover, the number of Ki67-positive, actively proliferative myCAFs also decreased in tumors derived from Areg -deficient EC1 cells (Figures S7G-H). These data suggest that tumor-cell-derived amphiregulin is a crucial factor in promoting myCAF proliferation in EC tumors. Given the established role of myCAFs in suppressing T cell function in PDAC, 36 , 37 we examined the relationship between myCAF abundance and immune infiltration in human tumors. In the TCGA-PAAD cohort, the myCAF signature score, defined as the geometric mean expression of established myCAF marker genes (Supplementary Table 2), was negatively correlated with the infiltration of multiple cytotoxic immune cell populations, including CD4+ naive T cells, CD8+ T cells, Th1 cells, and natural killer T (NKT) cells (Figure S7I). To test whether cancer cell-derived amphiregulin contributes to immune evasion, we examined T cell infiltration in tumors derived from Areg -deficient EC1 cells. Consistent with reduced myCAF abundance, genetic ablation of Areg significantly increased CD3+ and CD8+ T cell infiltration ( Figures 4J-M ). Notably, this effect persisted even in late-stage tumors (21 days post-orthotopic injection, as shown in Figures 4J-M ), when both EC and HSR tumors typically exhibit profound immune evasion. Collectively, these data demonstrate that Kras super-expressors shape an immunoevasive TME by delivering amphiregulin, which promotes myCAF expansion and suppresses T cell infiltration. Clonal expansion of KRAS -high cells fosters an immunoevasive spatial niche During PDAC progression in KPfC mice, we observed sporadic clusters of cancer cells exhibiting high Kras DNA FISH signals with morphological features resembling ecDNA (Figure S8). This finding suggests a multi-clonal emergence and expansion of Kras ecDNA-driven cancer cells, leading us to investigate whether clonal expansion of KRAS -high cells contributes to the formation of a spatially organized immunoevasive niche in human samples. To dissect the spatial organization of KRAS -high and KRAS -low cancer cell populations and their associated tumor microenvironments at single-cell resolution, we employed the Xenium in situ platform for high-plex spatial transcriptomics. We designed a custom 480-gene panel composed of established markers of cancer cells, immune cells, and CAF subtypes (Supplementary Table 3). We used it to profile six human PDAC specimens. In one specimen, a distinctive, heterogeneous KRAS expression pattern in cancer cells was identified (Figure S9A). Unsupervised clustering of this specimen identified the following major cell types: cancer cells, myCAFs, apCAFs, myeloid cells, T cells, endothelial cells, and mast cells ( Figures 5A-B , Figure S9B). Download figure Open in new tab Figure 5. Spatial analysis of KRAS -high and KRAS -low niches in PDAC. (A) UMAP visualization of major cell populations identified by Xenium spatial transcriptomics in a human PDAC sample. (B) UMAP visualization of EPCAM , KRAS , AREG, and CD3E expression in the human PDAC tissue of the Xenium spatial transcriptomic data. (C) Spatial distribution of major cell populations in representative tumor regions enriched for KRAS -high versus KRAS -low cancer cells. Cell types are color-coded as indicated. Number of regions: KRAS -high, n=6; KRAS -low, n=6. Scale bar: 50 μm. (D) Box plots displaying the proportional distribution of cell populations in spatial regions of KRAS -high and KRAS -low cancer cells. Number of regions: KRAS -high, n=6; KRAS -low, n=6. Wilcoxon test. We found that cancer cells were divided into two distinct clusters, characterized by differential KRAS expression levels ( Figure 5A , Figure S9B). Notably, KRAS -high cancer cells also exhibited elevated levels of AREG , recapitulating the results observed in mouse tumors ( Figure 5B , Figure S9B). Spatial visualization confirmed that AREG -expressing cells were predominantly localized within KRAS -high cancer cell regions (Figure S9C). To characterize the stromal composition of these distinct tumor subpopulations, we quantified CAF and immune cell densities in spatially defined regions enriched for either KRAS -high or KRAS -low cancer cells (Figure S9A). Consistently, KRAS -high regions exhibited significantly higher myCAF abundance and reduced T cell infiltration compared to KRAS -low regions ( Figures 5C-D , Figure S9C). These data demonstrate that the KRAS-AREG-axis mediated immunoevasive program, defined by myCAF enrichment and T cell exclusion, manifests in human PDAC as spatially distinct tumor microenvironments, driven by cancer cell-intrinsic heterogeneous KRAS expression. Discussion Alterations in oncogenes and tumor suppressor genes are central to TME remodeling 1 . Although ecDNA amplification has been associated with immune evasion 12 – 15 , 18 , its causal role remains unclear. Here, we show that ecDNA-driven amplification of Kras generates a small subpopulation of cancer cells with extremely high copy numbers. These Kras super-expressors downregulate immune response pathways and secrete amphiregulin (AREG), which establishes an immunoevasive niche characterized by enriched myCAF and reduced T cell infiltration. Although we do not exclude the possibility that a pre-existing cold TME favors ecDNA formation, our data provide direct evidence that ecDNA-mediated heterogeneity actively shapes the TME. Acentric ecDNA drives intratumoral heterogeneity through non-Mendelian segregation, producing daughter cells with variable ecDNA copy numbers. 9 , 38 This contributes to therapy resistance and poor prognosis. 10 , 11 , 39 – 41 Our study advances current understanding by demonstrating that, in the KPfC mouse PDAC model, ecDNA-driven Kras super-expressors act as powerful architects of an immunoevasive TME. In human PDAC, heterogeneous KRAS expression similarly correlates with spatial variation in myCAF and immune cell organization, where cancer cells with elevated KRAS levels are spatially associated with a more immunoevasive niche. However, excessive oncogene signaling, such as hyperactivation of the KRAS-ERK pathway, can impair cell fitness via senescence and apoptosis. 42 Signature analysis of our scRNAseq data reveals elevated cellular stress, including replication stress, in Kras super-expressors, explaining why ecDNA copy number cannot increase indefinitely. Notably, ecDNA enables rapid adaptation through asymmetric segregation, allowing daughter cells to resolve oncogene overdose more easily than chromosomal counterparts. Comparative analysis of ecDNA and HSR-driven expression reveals greater variation in Kras ecDNAs, but not in Myc ecDNAs. The weaker correlation between Myc ecDNA copy number and gene expression suggests additional regulatory mechanisms modulate Myc transcription, indicating that ecDNA copy number heterogeneity does not always translate into transcriptional heterogeneity. Furthermore, PDAC cells rarely harbor extremely high copies of both Kras and Myc ecDNAs, suggesting the presence of post-segregation selection pressures that refine ecDNA composition beyond co-segregation. 43 Myofibroblastic CAFs represent a critical barrier to anti-tumor immunity in PDAC. 36 , 44 , 45 These activated fibroblasts not only generate dense extracellular matrices that physically exclude immune cells from the tumor parenchyma, but also actively secrete immunosuppressive factors that impair T cell function. The abundance of myCAFs has been reported to be inversely associated with anti-tumor immunity, 45 underscoring the importance of understanding the molecular mechanisms that drive myCAF expansion in PDAC. Our study reveals that ecDNA-driven KRAS super-expressors actively remodel the TME by promoting myCAF proliferation, thereby establishing an immunoevasive niche that facilitates tumor progression. Integrating scRNAseq, spatial transcriptomics, and functional genetic data, we identify AREG as a critical mediator of KRAS super-expressor-driven myCAF expansion and subsequent immune exclusion. Previous studies have demonstrated that autocrine AREG signaling in myCAFs promotes metastasis by activating EGFR-mediated motility and invasiveness. 35 Our findings reveal a complementary paracrine mechanism in which cancer cells, specifically those with extremely high KRAS expression driven by ecDNA heterogeneity, secrete elevated levels of AREG to promote myCAF proliferation in the surrounding stroma, thereby excluding T cells. These data position AREG as a critical regulator during PDAC progression. While AREG emerges as a central player, other effector genes, including CEACAM1 , CSF2 (GM-CSF), PDCD1LG2 (PD-L2), and PTHLH , have also been implicated in PDAC pathology. 6 , 46 – 48 A comprehensive mapping of the ecDNA-mediated crosstalk landscape is required to fully understand the interaction of ecDNA-driven cancer cells and TME. In addition, the mechanisms underlying Kras and Myc ecDNA formation in PDAC remain unclear. The clonal organization of Kras -amplified cancer cells suggests that ecDNA genesis may require additional genetic events beyond Trp53 deficiency. 13 Future study is warranted to identify these prerequisites for ecDNA pathogenesis and their roles in tumor evolution. In summary, our findings highlight ecDNA as a dynamic genetic element that not only amplifies oncogene expression but also actively remodels the tumor ecosystem to promote immune evasion. This underscores the importance of incorporating ecDNA biology into therapeutic strategies targeting both cancer cell-intrinsic and microenvironmental vulnerabilities. STAR Methods Key resources table View this table: View inline View popup Experimental model and study participant details Mice KPfC ( Kras LSL-G12D/+ ; Trp53 fl/fl ; Pdx1 Cre/+ ) mice were generated as previously described. 22 The KPfC mouse had a pure C57BL/6 genetic background. All other mice used in this study were purchased from The Jackson Laboratory (6–9-week-old female C57BL/6J, strain #000664). All mice were maintained under specific pathogen-free conditions in the animal facility at the University of Texas Southwestern Medical Center at Dallas, in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC). Cell lines CT1BA5 cells were established as previously described from a female KPfC mouse PDAC model ( Kras LSL-G12D/+ ; Trp53 fl/fl ; Pdx1 Cre/+ ) on a C57BL/6 background. 23 Cells were cultured in Dulbecco’s Modified Eagle’s Medium/F12 (DMEM/F12; Corning, 10-092-CV) supplemented with 10% Fetal Bovine Serum (FBS; Corning, 35-011-CV) and 1× Penicillin-Streptomycin (Corning, 30-002-CI) at 37°C with 5% CO 2 . All cells were regularly tested for mycoplasma. Patient samples Human PDAC specimens were procured from the UT Southwestern Tissue Management Shared Resource following informed patient consent and approval by the UT Southwestern Institutional Review Board. Formalin-fixed paraffin-embedded (FFPE) tissue blocks were sectioned at 5-μm thickness. Histopathological diagnosis was verified by an experienced pathologist through hematoxylin and eosin staining of representative sections. Adjacent serial sections from these blocks were subsequently utilized for spatial transcriptomic profiling. Patient demographic information, including age and sex, was not disclosed to maintain confidentiality. Method details Tissue dissociation Tumor samples were minced into 1-2 mm 3 pieces on ice and transferred to 15 mL tubes containing warmed 10 mL tumor digestion buffer (9 mL DMEM + 1 mL 10× digestive buffer; see recipe below), then incubated at 37°C with shaking at 100 rpm for 30 minutes. The resulting suspensions were filtered through a 70-µm cell strainer, and enzymatic activity was quenched by adding ice-cold medium containing 10% fetal bovine serum (FBS). The suspensions were spun down and washed once using ice-cold medium with 10% FBS. To lyse blood cells, the cell pellet was resuspended gently by adding 5 mL ice-cold ACK lysing buffer (Thermo Fisher Scientific, A1049201) and incubated for 10 min on ice. Then ice-cold DPBS (Corning, 21031CV) with 0.04% BSA was added to the cells, and the cells were spun down at 300 ×g for 5 minutes at 4°C. The 10× digestive stock buffer contains collagenase type I (450 units/ml, Worthington, LS004214), collagenase type II (150 units/mL, Worthington, LS004202), collagenase type III (450 units/mL, Worthington, LS004206), collagenase type IV (450 units/mL, Worthington, LS004210), elastase (0.8 units/mL, Worthington, LS002290), hyaluronidase (300 units/mL, Sigma, H3506100MG), and DNase type I (250 units/mL, Sigma-Aldrich, D452710KU) in PBS. Metaphase chromosome spread Cells were treated with KaryoMAX (Gibco, 15212012) for 3 hours, and the single-cell suspension was harvested by trypsinization and centrifugation at 500 ×g for 5 minutes. After washing with PBS, cells were resuspended in 600 µL of 75 mM KCl and incubated at 37°C for 25 minutes, followed by fixation with 600 µL freshly prepared Carnoy’s fixative solution (3:1 Methanol: Glacial acetic acid). Cells were spun down at 800 ×g for 2 minutes. After another three fixations, cells were resuspended in 300 µl fixative solution, dropped onto the humidified slide, and air-dried for DAPI staining at room temperature. Finally, the slides were mounted with antifade mounting medium (Vector Laboratories, H-1000-10) and sealed with nail polish. OligoPaint DNA FISH on metaphase chromosome spreads Air-dried slides with metaphase chromosome spreads were equilibrated in 2× SSC and dehydrated in 70%, 85%, and 100% ethanol for 2 minutes each. Primary OligoPaint Kras and Myc probes diluted in hybridization buffer (1:4) were applied to slides and mounted with coverslips. Slides were denatured for 2 minutes at 75°C and hybridized overnight at 37°C in a humidified chamber. Next, the slides were washed twice with 0.4× SSC containing 0.3% IGEPAL for 10 minutes at 45°C, followed by a single wash with 2× SSC containing 0.1% IGEPAL for 10 minutes at room temperature. Slides were then carefully dried before applying the secondary OligoPaint probe labeled with a fluorophore, diluted in hybridization buffer (1:4), and incubating at 37°C for 1 hour in the dark. Post-hybridization washes were repeated as above. Finally, the slides were stained with DAPI and mounted using mounting medium (Prolong Diamond Antifade Mountant, Thermo Fisher Scientific, P36961). OligoPaint pool was designed and synthesized as previously reported. 49 In brief, approximately 1,000 oligonucleotides, each 68 to 80 bases long, were designed to span a 200-kb genomic region using OligoMiner software. Diverse two 20-nt adapters were added to the 5’ and 3’ ends to enable PCR, in vitro transcription, reverse transcription, and fluorescent secondary probe binding. The following genomic coordinates were used for OligoPaint pool design: Mouse Myc , chr15:61,950,801-62,150,800; Mouse Kras , chr6: 45,129,701-145,329,700. Whole genome sequencing (WGS) WGS libraries were prepared using NEBNext Ultra II DNA Library Prep Kit for Illumina (NEB, E7645L) following the manufacturer’s manual. Sequencing was performed in paired-end 150 bp (PE150) mode. Raw sequencing data were aligned to the mm10 mouse reference genome using bwa-mem2 (v2.2.1). 50 Aligned reads were sorted and deduplicated with sambamba (v1.0.1), 51 then analyzed with AmpliconArchitect (v1.5.r0) to extract ecDNA structure and copy number. 52 The Kras G12D mutation frequency was analyzed with the mpileup function in samtools (v1.21). 53 Bulk-cell RNA sequencing (RNAseq) RNAseq libraries were prepared using the NEBNext Ultra II RNA Library Prep Kit for Illumina (NEB, E7770L) with the NEBNext Poly(A) mRNA Magnetic Isolation Module (NEB, E7490) following the manufacturer’s manual. Sequencing was performed in paired-end 150 bp (PE150) mode. Raw sequencing data were analyzed using kallisto (v0.51.1) with the mm39 mouse reference genome. 54 Differentially expressed genes were identified using DESeq2. 55 Genes were considered significantly upregulated if they exhibited a log 2 fold-change ≥ 1 and a false discovery rate (FDR) < 0.05, and significantly downregulated if they exhibited a log 2 fold-change ≤ -1 and FDR < 0.05. Cell viability assay Cell viability was assessed using the Cell Counting Kit-8 (ApexBio Technology, K1018) following the manufacturer’s protocol. Briefly, 2×10 3 cells were seeded onto a 96-well plate. At each time point, 10 μL of CCK-8 reagent was added to wells containing 100 μL of cell culture medium, and the mixture was incubated for 2 hours at 37°C in a humidified incubator. Absorbance at 450 nm was recorded using the Infinite M200 Plex microplate reader (Tecan) as a surrogate for cell viability. Subcutaneous tumor implantation Single-cell suspensions were harvested by trypsinization, washed with PBS, and resuspended in DPBS (Corning, 21031CV) with 1 g/L glucose to achieve a final concentration of 3×10 6 cells/mL. A total of 50 μL of cells per mouse was injected subcutaneously into the right flank of an 8-week-old female mouse, and tumor size was monitored with a caliper every 3 days. Mice were euthanized by CO 2 inhalation once the tumor reached 2 cm in any dimension according to IACUC instructions. The tumor size was calculated using the formula: Volume = (π/6) × Length × Width × Height. Orthotopic tumor implantation Single-cell suspensions were collected as above to achieve a final concentration of 2.5×10 6 cells/mL. A total of 20 μL of cell suspension per mouse was injected orthotopically into the pancreas of female mice aged 6–9 weeks. Mice were euthanized by CO 2 inhalation at the same endpoint for tumor size assessment. For survival analysis, mice were euthanized when signs of distress were observed, and the time to death or euthanasia was used as the survival endpoint. Sample and library preparation for single-cell RNA sequencing (scRNAseq) Tumor-bearing mice were sacrificed on day 15 after cell inoculation. Pancreatic tumors with spleens from 12 mice were dissected, weighed, and rinsed. To avoid cell contamination from the normal pancreas and spleen, half of the tumor from each mouse was isolated from the opposite side of the spleen. To reduce individual variance among mice, two bisected tumors within each group were randomly combined. A total of 3 samples for each group were dissociated to prepare single-cell suspensions. Tissue digestion is identical to the method described in the tissue dissociation section above. To achieve greater than 85% cell viability, we used a dead-cell removal kit (Miltenyi, 130-090-101) to remove dead cells and assess cell viability with trypan blue staining. scRNAseq library preparation was performed according to the manufacturer’s guidelines (10x Genomics, user guide #CG000315 Rev E). Roughly 1×10 4 cells were loaded into the microfluidic system to generate a gel bead-in-emulsion. More than 2×10 4 read pairs per cell were captured. Sequencer-generated BCL files were converted to fastq format and subsequently aligned to the mm10 mouse genome using Cell Ranger (v7.2.0, 10x Genomics). Filtered output from Cell Ranger served as the basis for downstream analytical procedures. Data processing of scRNAseq The count matrices produced by Cell Ranger were loaded using the Read10X function from the Seurat package (v5.0.1) 56 and subsequently converted into Seurat objects. Quality control filtering was based on gene detection thresholds and mitochondrial content: cells with fewer than 300 or more than 7,000 detected genes, or cell exhibiting mitochondrial UMI counts exceeding 10%, were excluded. Genes detected in fewer than three cells were also removed. Potential doublets were identified and filtered out using DoubletFinder (v2.0.3) 57 with default parameters. Data normalization employed the NormalizeData function to standardize total expression across cells, followed by identification of the top 2,000 highly variable genes using TheFindVariableFeatures. The Seurat objects were scaled with ScaleData, incorporating regression of cell cycle effects based on phase scores calculated from known markers. 58 Dimensionality reduction was performed using the RunPCA function to conduct principal component analysis. Subsequently, a shared nearest neighbor (SNN) graph was built based on selected principal components using the FindNeighbors and FindClusters functions from the Seurat package. Cell clustering was performed using a graph-based approach with the Louvain algorithm. Differentially expressed marker genes for each cluster were identified using the FindAllMarkers and FindMarkers functions. Cell identity for each cluster was determined using a two-step approach that integrated marker-based and reference-based annotations. Initially, major cell types were manually assigned based on well-characterized canonical markers, including cancer cell markers ( Kras , Myc , Pvt1 , Krt18 ), immune cell markers ( Ptprc ), fibroblast markers ( Col1a1 , Col1a2 ), and endothelial cell markers ( Pecam1 , Emcn ). Ptprc -positive cells were further annotated using the SingleR package 24 with a reference dataset (main label). Identities assigned by SingleR were subsequently validated through canonical markers for macrophages ( Adgre1 , Arg1 , Itgax ), granulocytes ( S100a9 , S100a8 , G0s2 ), monocytes ( Itgam , Cd14 ), B cells ( Cd19 , Cd79b ), NK cells ( Nkg7 ), and T cells ( Cd3e , Cd3d ). Visualization was performed using the scCustomize package. 59 Subclustering followed comparable procedures, including normalization, variable gene detection, dimensionality reduction, and clustering. Cancer-associated fibroblast (CAF) subsets were annotated based on canonical markers distinguishing myCAFs ( Postn , Lrrc15 , Mmp11 , Acta2 , Thy1 ), mesothelial cells ( Msln , Upk3b , Krt19 , Lrrn4 ), iCAFs ( Dpt , Pi16 , Il6 , Cxcl12 , Cxcl1 ), proliferating CAFs ( Tpx2 , Ccna2 , Cenpf , Cdc20 , Mki67 ), and antigen-presenting CAFs ( H2 - Aa , H2 - Ab1 , Cd74 ). Immunohistochemistry and analysis All reagents used in immunohistochemistry were purchased from VECTASTAIN Elite ABC HRP Kit (Vector Laboratories, PK-6101) unless otherwise stated. Tumor samples were harvested after mice were euthanized by CO 2 inhalation and fixed with 4% paraformaldehyde in 0.1 M phosphate buffer (FD NeuroTechnologies, PF101) at 4°C for 24 hours and maintained in 70% ethanol solution at 4°C. Tissue sectioning was performed by the Tissue Management Shared Resource at the University of Texas Southwestern Medical Center. Sections of formalin-fixed, paraffin-embedded tissue were deparaffinized with Formula 83 (CBG Biotech, CH0104A) and rehydrated with graded alcohol. Antigen retrieval was performed using the Antigen Unmasking Solution (Vector Laboratories, H3300250) in a pressure cooker under 7.5 psi for 17 minutes. Slides were cooled down to room temperature, rinsed with Milli-Q water, and outlined with a PAP pen. Slides were blocked with BLOXALL Blocking Solution (Vector Laboratories, SP-6000) at room temperature for 10 minutes and washed with PBS. Slides were blocked with 1.5% normal goat serum in PBS at room temperature for 1 hour. The following primary antibodies were applied overnight at 4 °C: anti-POSTN (1:2,000, Cell Signaling Technology, 35126S), anti-CD3E (1:500, Thermo Fisher Scientific, MA5-35204), and anti-CD8A (1:500, Cell Signaling Technology, 98941S). Slides were washed with PBST and incubated with biotinylated secondary antibody, followed by sequential application of ABC Reagent. The DAB Substrate Kit (Vector Laboratories, SK-4105) was used for visualization. Finally, slides were counterstained with hematoxylin (Vector Laboratories, H-3401-500) and mounted with mounting medium (Vector Laboratories, H-5501-60). After drying at room temperature for 24 hours, slides were scanned by the Nanozoomer S60. To quantify DAB staining intensity, four images (each 0.715 mm²) were acquired from different regions of each sample using NDP viewer (v2.9.29). The staining intensity was defined as the ratio of DAB-positive area to the total tissue area. Image processing was performed using CellProfiler (v4.2.8). Briefly, original colored images were converted to grayscale, and the threshold was adjusted to segment DAB-positive regions and blank areas. Blank areas were excluded by subtracting non-tissue regions from the total image area. Immunofluorescence on FFPE sections Slides were treated in the same way as in immunohistochemistry experiments, from deparaffinization to antigen retrieval. Then, samples were blocked with 1.5% normal goat serum in PBS. The following antibodies were incubated overnight at 4°C: anti-RAS (1:200, Cell Signaling Technology, 91054), anti-MYC (1:100, Abcam, ab32072), anti-POSTN (1:2000, Cell Signaling Technology, 35126S), and anti-KI67 (1:100, Thermo Fisher Scientific, 14-5698-80). After washing with PBST, the following secondary antibodies were applied at room temperature for 1 hour: anti-Rabbit antibody conjugated with AlexaFluor 594 (1:1,000, Thermo Fisher Scientific, A32740), anti-Rabbit antibody conjugated with AlexaFluor 488 (1:1,000, Thermo Fisher Scientific, A32731), anti-Mouse antibody conjugated with AlexaFluor 594 (1:1,000, Thermo Fisher Scientific, A32742), and anti-Rat antibody conjugated with AlexaFluor 568 (1:1,000, Thermo Fisher Scientific, A11077). After washing with PBST buffer, autofluorescence was quenched using TrueVIEW reagents (Vector Laboratories, SP-8400-15). Sections were counterstained with DAPI and mounted with antifade medium (Thermo Fisher Scientific, P36961). The MYC protein fluorescence intensity was quantified using CellProfiler (v4.2.8). Briefly, nuclei were identified based on the DAPI channel, and the resulting nuclear masks were applied to the MYC protein fluorescence channels. The fluorescence intensity of the MYC signal within each nucleus was measured. The RAS protein fluorescence intensity was quantified in ImageJ by manually outlining each cell and measuring the fluorescence intensity in the RAS signal channel. OligoPaint DNA FISH on FFPE sections DNA FISH on formalin-fixed paraffin-embedded (FFPE) tissue sections was carried out following the steps outlined in our previously published protocol. 60 Briefly, FFPE sections were deparaffinized with Formula 83 and rehydrated with an ethanol series. Tissues were then treated with 0.2N HCl at room temperature for 20 minutes, 10 mM citric acid at 90°C for 20 minutes, and proteinase K at room temperature for 1 minute. After dehydration with an ethanol series, OligoPaint FISH was performed as described above. Finally, autofluorescence was eliminated using the TrueVIEW Autofluorescence Quenching Kit (Vector Laboratories, SP-8400) before DAPI counterstaining. The FISH signal intensity was quantified using CellProfiler (v4.2.8). Briefly, nuclei were identified based on the DAPI channel, and the resulting nuclear masks were applied to the FISH signal channels. The FISH signals were refined into sharp puncta by threshold adjustment, and the total signal intensity within each nucleus was measured. To combine OligoPaint DNA FISH and IF in the same section, OligoPaint DNA FISH was first performed as described above. After secondary probe hybridization and washing, slides were blocked with 10% normal goat serum with 0.05% Triton X-100 in PBS for 1 hour, followed by immunostaining as described in the immunofluorescence section. Fluorescence microscopy DNA FISH and immunofluorescence imaging were captured using a Zeiss Axio Observer 7 microscope equipped with the Apotome 3 optical sectioning system. Images were acquired with a 63× Plan-Apochromat oil immersion objective lens (NA 1.40) and processed using Zeiss ZEN software (version 3.4). For each sample, at least five z-stack images were captured at 1 μm intervals, followed by the generation of maximum intensity projections to produce 2D representations. Whole tissue scans of FISH or combined FISH and immunofluorescence staining were performed using a Zeiss Axioscan 7 slide scanner with a 40× objective. Gene signature, immune infiltration analysis, and survival analysis Kras super-expressor signature genes (Supplementary Table 1) were identified using Seurat (v5) based on the following criteria: log 2 fold-change ≥ 1.5, presence of human orthologs, and ranking among the top 30 genes by FDR. These genes were mapped to their human orthologs to retrieve transcript per million (TPM) values from the TCGA-PAAD cohort, obtained via UCSC Xena. Kras super-expressor signature scores were calculated as the geometric mean of TPM values for the corresponding human orthologs. Similarly, myCAF scores were defined as the geometric mean of TPM values for a curated list of myCAF biomarker genes (Supplementary Table 2). Immune infiltration scores were sourced from TIMEDB. Pearson correlation analysis was performed to assess association significance, and p-values were adjusted using the FDR method to control for false positives. Overall survival analysis was performed using GEPIA2 (gepia2.cancer-pku.cn). 61 The cutoff for distinguishing low and high expression was set as 50% of the median. RNAi Kras siRNA (Thermo Fisher Scientific, s68936) was delivered to cells using Lipofectamine RNAiMAX (Thermo Fisher Scientific, 13778150) according to the manufacturer’s protocol. Transfected cells were harvested 48 hours after transfection for subsequent experimental analyses. Western blot Cells treated with inhibitors were harvested and lysed in RIPA buffer (Boston BioProducts, BP-115) with protease inhibitor (Roche, 04693116001) and phosphatase inhibitor (Roche, 04906837001). Protein concentration was determined using the BCA Protein Assay Kit (Thermo Fisher Scientific, A53225). Equal amount of protein samples were prepared in 1× Laemmli sample buffer (Bio-Rad, 1610747) and heated at 95°C for 10 minutes, ran on 4-20% Mini-PROTEAN gels (Bio-Rad, 4568096), and transferred to a nitrocellulose membrane using Trans-Blot Turbo system with transfer kit (Bio-Rad, 1704270). The membrane was blocked with 5% BSA in TBS with 0.1% Tween-20 (Fisher Scientific, BP337-500) for 1 hour. The following primary antibodies were used: anti-pMAPK (1:2,000, Cell Signaling Technology, 4370T), anti-MAPK (1:1000, Cell Signaling Technology, 4695T), anti-p4E-BP1 (1:1,000, Cell Signaling Technology, 2855T), anti-4E-BP1 (1:1,000, Cell Signaling Technology, 9644T), and anti-Vinculin (1:10,000, Proteintech, 66305-1-Ig). The membranes were washed 3 times with TBST and incubated with HRP-conjugated secondary antibodies (Anti-mouse, 1:5,000, Cell Signaling Technology, 7076S; Anti-rabbit, 1:5,000, Cell Signaling Technology, 7074S) for 1 hour. Membranes were washed 3 times with TBST, then visualized with a chemiluminescent substrate (Thermo Fisher Scientific, 34580) and imaged using ImageQuant 800 (Amersham). RT-qPCR RNA was extracted using the Quick-RNA Miniprep Plus Kit (Zymo Research, 50-125-1683) and quantified by NanoDrop (Thermo Scientific). One microgram of RNA was used for reverse transcription to generate cDNA using the Maxima H Minus reverse transcriptase (Thermo Scientific, EP0753). Primers used are in the table of the method section. The cDNA was added with 1× SYBR Green qPCR Master Mix (Selleck Chemicals, B21203) and 400 nM of the respective forward and reverse primers for qPCR on a CFX Opus 96 System (Bio-Rad). The relative RNA level was calculated by the ddCt method. CRISPR knockout Two sgRNAs were designed using E-CRISP, 62 synthesized by Integrated DNA Technologies, and cloned into the LentiCRISPRv2 vector (Addgene, 52961). Lentivirus packaging was performed by co-transfecting HEK293T cells with lentiviral transfer plasmids and helper plasmids (pMD2.G, Addgene, 12259; psPAX2, Addgene, 12260). The CT1BA5EC cells were infected with the lentivirus and selected by puromycin (InvivoGen, ant-pr-1). Single-cell clones were isolated using a flow cytometer (BD FACSAria Fusion) and genotyped to confirm homozygous knockout. Xenium in situ high-plex assay We developed a fully customized 480-gene panel (Supplementary Table 3) for cell type identification and functional characterization. FFPE tissue blocks were sectioned at 5-μm thickness and mounted within the designated capture area (10.45 mm × 22.45 mm) of Xenium slides (10x Genomics). Sample preparation followed the manufacturer’s protocol for FFPE tissues (10x Genomics Xenium In Situ FFPE Tissue Preparation Guide). Mounted sections underwent deparaffinization in xylene followed by gradient rehydration through graded ethanol washes. Enzymatic permeabilization was then performed to enhance mRNA accessibility. Target mRNA hybridization was conducted overnight at 50°C using gene-specific probes, after which excess unbound probes were washed away. Locked nucleic acid probes annealed to target transcripts were subsequently ligated using Xenium ligase A/B at 37°C for 2 hours. Gene-specific circular probes were then amplified by rolling-circle amplification to generate multiple copies of unique molecular barcodes. Following additional wash steps, autofluorescence was eliminated through chemical quenching. Imaging was performed on a Xenium analyzer (10x Genomics), with DAPI counterstaining providing nuclear morphology for cellular boundaries. Cell segmentation was accomplished using the instrument’s integrated machine learning algorithms, and the resulting data files were processed for subsequent computational analysis. Xenium data processing and analysis Xenium spatial transcriptomics data were analyzed using Seurat (v5.3.0) following standard workflows with modifications. Cell-by-gene expression matrices generated by the Xenium analyzer were imported into R and consolidated into a unified Seurat object. Quality control filtering was applied to remove low-quality cells, followed by variance-stabilizing transformation and normalization via the SCTransform method. For unsupervised cell type identification, we performed dimensionality reduction via principal component analysis on highly variable genes (RunPCA). A k-nearest neighbor graph was constructed in the reduced-dimensional PCA space (FindNeighbors), followed by graph-based clustering using the Louvain algorithm (FindClusters, resolution = 0.2). Cell type identities were assigned to each cluster based on the expression of canonical marker genes. Clustering results were visualized in both reduced-dimensional space via Uniform Manifold Approximation and Projection (UMAP) and in their original spatial coordinates to preserve tissue architecture. Statistics All experiments were performed for at least three times. Statistical analyses were detailed in the corresponding figure legends. For data that were normally distributed and exhibited homoscedasticity, Student’s t-test or ANOVA was used. Otherwise, the corresponding statistical analysis was indicated in the figure legend. P values derived from multiple comparisons were adjusted using the false discovery rate (FDR) method. All tests were two-sided. Box plots display the median and interquartile range (IQR), with whiskers extending to 1.5× the IQR. Data points and error bars shown on line plots represent the mean ± standard error of the mean (SEM). Data availability All sequencing data will be uploaded to the NIH Sequence Read Archive repository and made publicly accessible upon the acceptance of this manuscript. Author contributions Kailiang Qiao: Investigation, Validation, Formal analysis, Visualization, Writing - Review & Editing. Qing-Lin Yang: Investigation, Validation, Formal analysis, Visualization, Writing - Review & Editing. Tuo Li: Investigation, Writing - Review & Editing. Xiongfeng Chen: Formal analysis. Zeynep Yazgan: Investigation. Yoon Jung Kim: Investigation. Collin Gilbreath: Investigation. Jun Yi Stanley Lim: Investigation. Yipeng Xie: Investigation. Xiaohui Sun: Investigation. Yang Liu: Investigation. Yiyue Jia: Investigation. Zhijian J. Chen: Conceptualization, Writing - Original Draft, Funding acquisition. Huocong Huang: Conceptualization, Methodology, Formal analysis, Writing - Original Draft, Funding acquisition. Sihan Wu: Conceptualization, Methodology, Resources, Formal analysis, Writing - Original Draft, Funding acquisition, Visualization, Supervision. Competing interests Sihan Wu is a member of the SAB of Dimension Genomics. Zhijian J. Chen is an Investigator of the Howard Hughes Medical Institute. The authors declare no competing financial interests. Supplementary Figure Legends Figure S1. Amplicon structure of EC and HSR PDAC cell lines. The structures of Kras and Myc amplicons in EC1/2 and HSR1/2 PDAC cell lines were analyzed using AmpliconArchitect, showing high similarity between ecDNA and HSR amplicons. Figure S2. Molecular characteristics of EC and HSR PDAC cell lines. (A) Stacked bar plot showing the percentage of KRAS G12D allele in EC1/2 and HSR1/2 isogenic pairs. (B) Principal component analysis (PCA) of bulk-cell gene expression profiles showing clear separation between EC and HSR cell lines. Rep1 and Rep2 indicate two sequencing replicates. (C) Heatmap showing that only 1% of genes in the whole transcriptome were differentially expressed among EC and HSR groups. (D) GSEA analysis indicating minimal differences for KRAS and MYC signaling between the EC and HSR transcriptome in bulk-cell cultured in vitro . (E) Cell viability assay in vitro using CCK-8. Data represent mean ± SEM from three independent experiments. Two-way ANOVA test with Tukey’s HSD. NS indicates not significant. Figure S3. Expression of canonical biomarkers of each cell population (A) Gross images and tumor weight quantification of tumors for scRNAseq. The mid-stage tumor weight between EC1 and HSR1 has not shown significant differences. Student’s t-test. (B) Dot plot showing expression of canonical marker genes across distinct cell populations to validate the cell type annotation identified by SingleR. Dot size indicates the percentage of cells expressing the gene within each cell type, and dot color represents the scaled expression level. (C) Dot plot showing expression of canonical marker genes across CAF subtypes to annotate the subclusters after dimension reduction. Dot size indicates the percentage of cells expressing the gene within each cell type, and dot color represents the scaled expression level. (D) UMAP visualization of T cell subclusters. (E) Dot plot showing expression of canonical T cell markers. (F) Cell composition analysis of T cell subtypes in EC and HSR groups. Fisher’s exact test. (G) Representative images of CD3E IHC staining for mid-stage tumors showing both low- and high-magnification views. Red letter S and circles indicate the spleen tissue areas as a positive control for CD3E staining. Scale bars 5 mm (low magnification), 50 μm (high magnification). Figure S4. Variance and correlation analyses among ecDNA copy number, mRNA transcription, and protein expression. (A) Representative Kras DNA FISH images taken from FFPE tumor sections. Scale bar: 10 μm. (B) Density plot showing the distribution of Kras DNA FISH signal intensity of FFPE tumor sections in each cell. The variances of EC and HSR were assessed using MAD and IQR. Number of nuclei analyzed: EC, n = 984; HSR, n = 776. (C) Representative RAS protein IF images taken from FFPE tumor sections. Scale bar: 10 μm. (D) Density plot showing the distribution of RAS protein IF intensity of tumor section in each cell. The variances of EC and HSR were assessed using MAD and IQR. Number of cells analyzed: EC, n = 288; HSR, n = 402. (E) Representative images of RAS protein IF and Kras DNA FISH co-staining in FFPE tumor sections. Scale bar: 10 μm. (F) Correlation analysis between the intensity of RAS protein IF and the intensity of Kras DNA FISH in FFPE tumor sections. Spearman correlation test. Number of cells analyzed: n = 723. (G) Density plot of Myc mRNA expression of EC and HSR cells from scRNAseq data. The distribution variance was assessed using MAD and IQR. Number of cells analyzed: EC, n = 10,000; HSR, n = 10,000. (H) Representative images of MYC protein IF and Myc DNA FISH co-staining in FFPE tumor sections. Scale bar: 10 μm. (I) Correlation analysis between the intensity of MYC protein IF and the intensity of Myc DNA FISH in FFPE tumor sections. Spearman correlation test. Number of cells analyzed: n = 977. Figure S5. Comparative analysis of Kras super-expressors and normal-expressors. (A) Schematic workflow for comparative analysis of Kras super-expressors and normal-expressors. (B) Density plot of Kras mRNA expression from 10,000 randomly selected cancer cells, each from EC and HSR tumors. Cells with the top 5% Kras expression were defined as Kras super-expressors. GSEA analysis showed that KRAS signaling was significantly upregulated in the super-expressor compared with the normal-expressor, but not between EC and HSR in the normal-expressor. (C) UMAP view displaying the Kras super-expressor cells in the EC and HSR groups. (D) Dot plot showing the expression of selected replication stress marker genes between Kras super-expressors and normal-expressors. (E) Correlation analysis between the intensity of Kras DNA FISH and the intensity of Myc DNA FISH in FFPE EC tumor sections. Spearman correlation test. Number of nuclei analyzed: n = 1284. (F) Schematic illustrating the inheritance patterns of ecDNA and chromosomes. Asymmetric mitotic segregation of ecDNA can rapidly generate oncogene super-expressors, promoting tumor growth. However, elevated oncogene dosage induces cellular stress and reduces fitness. The dynamic nature of ecDNA inheritance enables reversible transitions between super- and normal-expressors, enhancing cancer cell adaptability to the tumor microenvironment. In contrast, the stable inheritance of chromosomes restricts the emergence of super-expressors and limits their interconversion with normal-expressors. Figure S6. Effector gene analysis in Kras super-expressors. (A) Dot plot showing the differential expressions of CEACAM1 , PDCD1LG2 , PTHLH, and CSF2 between TCGA-PAAD tumor samples and GTEx normal tissues. Wilcoxon test. (B) Overall survival of TCGA-PAAD cohort stratified by CEACAM1 , PDCD1LG2 , PTHLH , and CSF2 expression (median 50% cutoff). Log-rank test. (C-D) Violin plot showing the Log1p normalized mRNA expression of Areg and Csf2 in super- and normal- expressors. Wilcoxon test with FDR correction. (E) Violin plot showing the Log1p normalized mRNA expression of Areg in EC and HSR cells. Wilcoxon test with FDR correction. (F) qPCR detection of Areg and Kras mRNA levels in EC cells with or without transfection of Kras siRNA and siRNA control. Student’s t-test. (G) Western plot showing the expression of MAPK and pMAPK (T202/Y204) in EC cells treated with trametinib and SCH772984 for 24 hours in a dose-dependent way. (H) Western plot showing the expression of 4E-BP and p4E-BP (T37/46) in EC cells treated with Rapamycin and Torin 2 for 24 hours in a dose-dependent way. (I) qPCR detection of Areg mRNA level in EC cells treated with vehicle control DMSO, SCH772984 (2 μM), Trametinib (40 nM), Rapamycin (0.125 μM), Torin2 (200 nM), and combination treatment with Trametinib (40 nM) and Torin2 (200 nM) for 24 hours. One-way ANOVA with Tukey’s HSD correction. (J) Schematic image showing the KRAS signaling pathways. Figure S7. Areg knockout validation and extended data for Figure 4 . (A) qPCR detection of Areg mRNA level in two Areg- knockout clones (sg Areg #1 and #2) compared to non-targeting sgRNA control (sgNTC). (B) Kras and Myc DNA FISH on metaphase spreads from Areg -knockout cells. Scale bar: 5 μm. (C-D) Kras and Myc ecDNA counts in sgNTC and sg Areg clones. One-way ANOVA revealed no significant difference. Sample sizes: sgNTC, n = 31; sg Areg #1, n = 35; sg Areg #2, n = 32. (E) In vitro cell viability of sgNTC and sg Areg clones measured by CCK-8. Two-way ANOVA computed p-values with Tukey’s HSD correction. (F) Log1p normalized Areg and Egfr mRNA expression among different cell populations detected by scRNAseq. (G) Representative images of POSTN and KI67 protein IF staining. Blue arrows indicate the KI67-negative myCAFs, and yellow arrows indicate KI67-positive myCAFs. Scale bar: 50 μm (low magnification), 20 μm (high magnification). (H) Percentage of the KI67-positive myCAFs among total myCAFs. One-way ANOVA test with Tukey’s HSD correction. More than 100 POSTN+ myCAFs per mouse were counted. The p-value indicates the significance compared sgNTC. Sample sizes: sgNTC, n = 7; sg Areg #1, n = 7; sg Areg #2, n = 7. (I) Pearson correlation analysis between the myCAF score (calculated as the geometric mean of TPMs of myCAF biomarker genes) and immune cell infiltration scores obtained from TIMEDB, across samples in the TCGA-PAAD cohort. Figure S8. Spatial organization of cells with Kras amplification in autochthonous KPfC tumor. Representative images of the whole slide scanning stained by Kras DNA FISH from an autochthonous KPfC mouse tumor. Figure S9. Spatial organization of regions with high- and low- KRAS expression in human PDAC. (A) Density map of KRAS expression in the entire section from one human PDAC tumor (left) and cell identity analyzed by Xenium spatial transcriptomics (right). The locations of the selected regions for cell density analysis, shown in Figures 5C-D , were marked by red ( KRAS -high) and yellow ( KRAS -low) squares. Scale bar: 1 mm. (B) Dot plot showing marker gene expression of major cell populations identified by Xenium spatial transcriptomics in a human PDAC sample. (C) Subcellular spatial mapping of transcripts for EPCAM , KRAS , and AREG in a representative region with KRAS -high or KRAS -low cancer cells. Scale bar: 50 μm. Acknowledgement Sihan Wu is supported by the Cancer Prevention and Research Institute of Texas (CPRIT, RR210034) and the American Cancer Society (CAT-24-1379043-01-CAT). Huocong Huang is supported by the National Cancer Institute (R00 CA252009). Jun Yi Stanley Lim is supported by the CPRIT Training Grant (RP210041). This work was delivered as part of the eDyNAmiC team supported by the Cancer Grand Challenges partnership funded by Cancer Research UK (P.S.M CGCATF-2021/100012, Z.J.C., S.W. CGCATF-2021/100023) and the National Cancer Institute (P.S.M. OT2CA278688, Z.J.C., S.W. OT2CA278683). Work in the Chen laboratory is supported by grants from the National Cancer Institute (R01CA299257) and the Welch Foundation (I-1389). We acknowledge the assistance of the University of Texas Southwestern Tissue Management Shared Resource, a shared resource at the Simmons Comprehensive Cancer Center, which is supported in part by the National Cancer Institute under award number P30 CA142543. We acknowledge the assistance of the University of Texas Southwestern Whole Brain Microscopy Facility, RRID: SCR_017949. Certain graphic elements were obtained from the open-sourced Bioicons and NIH BIOART (CC-BY 3.0 unported license). Funder Information Declared Cancer Prevention and Research Institute of Texas , RR210034 National Cancer Institute , R00CA252009 , R01CA299257 CPRIT Training Grant , RP210041 Cancer Grand Challenges funded by Cancer Research UK , CGCATF-2021/100012 , CGCATF-2021/100023 Cancer Grand Challenges funded by the National Cancer Institute , OT2CA278688 , OT2CA278683 Welch Foundation , I-1389 Footnotes ↵ # Co-first authors: Kailiang Qiao, Qing-Lin Yang References 1. ↵ van Weverwijk , A. & de Visser , K. E . Mechanisms driving the immunoregulatory function of cancer cells . Nat. Rev. Cancer 23 , 193 – 215 ( 2023 ). OpenUrl CrossRef PubMed 2. ↵ Binnewies , M. et al. Understanding the tumor immune microenvironment (TIME) for effective therapy . Nat. Med . 24 , 541 – 550 ( 2018 ). OpenUrl CrossRef PubMed 3. ↵ Casacuberta-Serra , S. , González-Larreategui , Í. , Capitán-Leo , D. & Soucek , L . MYC and KRAS cooperation: from historical challenges to therapeutic opportunities in cancer . Signal Transduct. Target. Ther . 9 , 205 ( 2024 ). OpenUrl PubMed 4. Donahue , K. L. et al. Oncogenic KRAS-dependent stromal interleukin-33 directs the pancreatic microenvironment to promote tumor growth . Cancer Discov . 14 , 1964 – 1989 ( 2024 ). OpenUrl CrossRef PubMed 5. Casey , S. C. , Baylot , V. & Felsher , D. W . MYC: Master regulator of immune privilege . Trends Immunol . 38 , 298 – 305 ( 2017 ). OpenUrl CrossRef PubMed 6. ↵ Bayne , L. J. et al. Tumor-derived granulocyte-macrophage colony-stimulating factor regulates myeloid inflammation and T cell immunity in pancreatic cancer . Cancer Cell 21 , 822 – 835 ( 2012 ). OpenUrl CrossRef PubMed Web of Science 7. ↵ Pylayeva-Gupta , Y. , Lee , K. E. , Hajdu , C. H. , Miller , G. & Bar-Sagi , D . Oncogenic Kras-induced GM-CSF production promotes the development of pancreatic neoplasia . Cancer Cell 21 , 836 – 847 ( 2012 ). OpenUrl CrossRef PubMed Web of Science 8. ↵ Wu , S. et al. Circular ecDNA promotes accessible chromatin and high oncogene expression . Nature 575 , 699 – 703 ( 2019 ). OpenUrl CrossRef PubMed 9. ↵ Wu , S. , Bafna , V. , Chang , H. Y. & Mischel , P. S . Extrachromosomal DNA: An emerging hallmark in human cancer . Annu. Rev. Pathol . 17 , 367 – 386 ( 2022 ). OpenUrl CrossRef PubMed 10. ↵ Kim , H. et al. Extrachromosomal DNA is associated with oncogene amplification and poor outcome across multiple cancers . Nat. Genet . 52 , 891 – 897 ( 2020 ). OpenUrl CrossRef PubMed 11. ↵ Pal Choudhuri , S. , et al. Acquired cross-resistance in small cell lung cancer due to extrachromosomal DNA amplification of MYC paralogs . Cancer Discov . 14 , 804 – 827 ( 2024 ). OpenUrl CrossRef PubMed 12. ↵ Wu , T. et al. Extrachromosomal DNA formation enables tumor immune escape potentially through regulating antigen presentation gene expression . Sci. Rep . 12 , 3590 ( 2022 ). OpenUrl PubMed 13. ↵ Lin , M. S. et al. Transcriptional immune suppression and up-regulation of double-stranded DNA damage and repair repertoires in ecDNA-containing tumors . Elife 12 , ( 2024 ). 14. ↵ Lv , W. et al. Spatial-temporal diversity of extrachromosomal DNA shapes urothelial carcinoma evolution and tumor-immune microenvironment . Cancer Discov . 15 , 1225 – 1246 ( 2025 ). OpenUrl 15. ↵ Fiorini , E. et al. MYC ecDNA promotes intratumour heterogeneity and plasticity in PDAC . Nature 640 , 811 – 820 ( 2025 ). OpenUrl PubMed 16. ↵ Luebeck , J. et al. Extrachromosomal DNA in the cancerous transformation of Barrett’s oesophagus . Nature 616 , 798 – 805 ( 2023 ). OpenUrl CrossRef PubMed 17. Pang , J. et al. Extrachromosomal DNA in HPV-mediated oropharyngeal cancer drives diverse oncogene transcription . Clin. Cancer Res . 27 , 6772 – 6786 ( 2021 ). OpenUrl Abstract / FREE Full Text 18. ↵ Bailey , C. et al. Origins and impact of extrachromosomal DNA . Nature 635 , 193 – 200 ( 2024 ). OpenUrl CrossRef PubMed 19. ↵ Pradella , D. et al. Engineered extrachromosomal oncogene amplifications promote tumorigenesis . Nature 637 , 955 – 964 ( 2025 ). OpenUrl CrossRef PubMed 20. ↵ Maddipati , R. & Stanger , B. Z . Pancreatic cancer metastases harbor evidence of polyclonality . Cancer Discov . 5 , 1086 – 1097 ( 2015 ). OpenUrl Abstract / FREE Full Text 21. Shi , Y. et al. Targeting LIF-mediated paracrine interaction for pancreatic cancer therapy and monitoring . Nature 569 , 131 – 135 ( 2019 ). OpenUrl CrossRef PubMed 22. ↵ Hosein , A. N. et al. Cellular heterogeneity during mouse pancreatic ductal adenocarcinoma progression at single-cell resolution . JCI Insight 5 , ( 2019 ). 23. ↵ Huang , H. et al. Targeting TGFβR2-mutant tumors exposes vulnerabilities to stromal TGFβ blockade in pancreatic cancer . EMBO molecular medicine vol. 11 e10515 ( 2019 ). OpenUrl 24. ↵ Aran , D. et al. Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage . Nat. Immunol . 20 , 163 – 172 ( 2019 ). OpenUrl CrossRef PubMed 25. ↵ Chen , X. et al. Single-cell resolution spatial analysis of antigen-presenting cancer-associated fibroblast niches . Cancer Cell 0 , ( 2025 ). 26. Huang , H. et al. Mesothelial cell-derived antigen-presenting cancer-associated fibroblasts induce expansion of regulatory T cells in pancreatic cancer . Cancer Cell 40 , 656 – 673 .e7 ( 2022 ). OpenUrl CrossRef PubMed 27. Elyada , E. et al. Cross-species single-cell analysis of pancreatic ductal adenocarcinoma reveals antigen-presenting cancer-associated fibroblasts . Cancer Discov . 9 , 1102 – 1123 ( 2019 ). OpenUrl Abstract / FREE Full Text 28. ↵ Gao , Y. et al. Cross-tissue human fibroblast atlas reveals myofibroblast subtypes with distinct roles in immune modulation . Cancer Cell 42 , 1764 – 1783 .e10 ( 2024 ). OpenUrl CrossRef PubMed 29. ↵ Tay , C. , Tanaka , A. & Sakaguchi , S . Tumor-infiltrating regulatory T cells as targets of cancer immunotherapy . Cancer Cell 41 , 450 – 465 ( 2023 ). OpenUrl CrossRef PubMed 30. ↵ Lange , J. T. et al. The evolutionary dynamics of extrachromosomal DNA in human cancers . Nat. Genet . 54 , 1527 – 1533 ( 2022 ). OpenUrl CrossRef PubMed 31. ↵ Bibby , J. A. et al. Systematic single-cell pathway analysis to characterize early T cell activation . Cell Rep . 41 , 111697 ( 2022 ). OpenUrl CrossRef PubMed 32. ↵ Shaw , K. , Bernards , R. , Stegmaier , K. , Varmus , H. & Sellers , W. R . Prospects for understanding and exploiting the consequences of hyperactivation lethality . Trends Cancer 11 , 619 – 628 ( 2025 ). OpenUrl PubMed 33. ↵ Wang , X. et al. TIMEDB: tumor immune micro-environment cell composition database with automatic analysis and interactive visualization . Nucleic Acids Res . 51 , D1417 – D1424 ( 2023 ). OpenUrl CrossRef PubMed 34. ↵ Jin , S. , Plikus , M. V. & Nie , Q . CellChat for systematic analysis of cell-cell communication from single-cell transcriptomics . Nat. Protoc . 20 , 180 – 219 ( 2025 ). OpenUrl CrossRef PubMed 35. ↵ Mucciolo , G. et al. EGFR-activated myofibroblasts promote metastasis of pancreatic cancer . Cancer Cell 42 , 101 – 118 .e11 ( 2024 ). OpenUrl CrossRef PubMed 36. ↵ Krishnamurty , A. T. et al. LRRC15+ myofibroblasts dictate the stromal setpoint to suppress tumour immunity . Nature 611 , 148 – 154 ( 2022 ). OpenUrl CrossRef PubMed 37. ↵ Arpinati , L. & Scherz-Shouval , R . From gatekeepers to providers: regulation of immune functions by cancer-associated fibroblasts . Trends Cancer 9 , 421 – 443 ( 2023 ). OpenUrl PubMed 38. ↵ Yang , Q.-L. , Xie , Y. , Qiao , K. , Lim , J. Y. S. & Wu , S . Modern biology of extrachromosomal DNA: A decade-long voyage of discovery . Cell Res . 35 , 11 – 22 ( 2025 ). OpenUrl PubMed 39. ↵ Pongor , L. S. et al. Extrachromosomal DNA amplification contributes to small cell lung cancer heterogeneity and is associated with worse outcomes . Cancer Discov . 13 , 928 – 949 ( 2023 ). OpenUrl CrossRef PubMed 40. Chapman , O. S. et al. Circular extrachromosomal DNA promotes tumor heterogeneity in high-risk medulloblastoma . Nat. Genet . 55 , 2189 – 2199 ( 2023 ). OpenUrl CrossRef PubMed 41. ↵ Song , K. et al. Plasticity of extrachromosomal and intrachromosomal BRAF amplifications in overcoming targeted therapy dosage challenges . Cancer Discov . 12 , 1046 – 1069 ( 2022 ). OpenUrl CrossRef PubMed 42. ↵ Unni , A. M. et al. Hyperactivation of ERK by multiple mechanisms is toxic to RTK-RAS mutation-driven lung adenocarcinoma cells . Elife 7 , ( 2018 ). 43. ↵ Hung , K. L. et al. Coordinated inheritance of extrachromosomal DNAs in cancer cells . Nature 635 , 201 – 209 ( 2024 ). OpenUrl CrossRef PubMed 44. ↵ Caligiuri , G. & Tuveson , D. A . Activated fibroblasts in cancer: Perspectives and challenges . Cancer Cell 41 , 434 – 449 ( 2023 ). OpenUrl CrossRef PubMed 45. ↵ Dominguez , C. X. et al. Single-cell RNA sequencing reveals stromal evolution into LRRC15+ myofibroblasts as a determinant of patient response to cancer immunotherapy . Cancer Discov . 10 , 232 – 253 ( 2020 ). OpenUrl Abstract / FREE Full Text 46. ↵ Zińczuk , J. et al. Expression of chosen carcinoembryonic-related cell adhesion molecules in pancreatic intraepithelial neoplasia (PanIN) associated with chronic pancreatitis and pancreatic ductal adenocarcinoma (PDAC) . Int. J. Med. Sci . 16 , 583 – 592 ( 2019 ). OpenUrl PubMed 47. Zhang , Y. et al. A PD-L2-based immune marker signature helps to predict survival in resected pancreatic ductal adenocarcinoma . J. Immunother. Cancer 7 , 233 ( 2019 ). 48. ↵ Pitarresi , J. R. et al. PTHrP drives pancreatic cancer growth and metastasis and reveals a new therapeutic vulnerability . Cancer Discov . 11 , 1774 – 1791 ( 2021 ). OpenUrl Abstract / FREE Full Text 49. ↵ Beliveau , B. J. et al. Versatile design and synthesis platform for visualizing genomes with Oligopaint FISH probes . Proc. Natl. Acad. Sci. U. S. A . 109 , 21301 – 21306 ( 2012 ). OpenUrl Abstract / FREE Full Text 50. ↵ Vasimuddin , M. , Misra , S. , Li , H. & Aluru , S. Efficient architecture-aware acceleration of BWA-MEM for multicore systems . in 2019 IEEE International Parallel and Distributed Processing Symposium (IPDPS) 314 – 324 (IEEE, 2019 ). 51. ↵ Tarasov , A. , Vilella , A. J. , Cuppen , E. , Nijman , I. J. & Prins , P . Sambamba: fast processing of NGS alignment formats . Bioinformatics 31 , 2032 – 2034 ( 2015 ). OpenUrl CrossRef PubMed 52. ↵ Deshpande , V. et al. Exploring the landscape of focal amplifications in cancer using AmpliconArchitect . Nat. Commun . 10 , 392 ( 2019 ). 53. ↵ Danecek , P. et al. Twelve years of SAMtools and BCFtools . Gigascience 10 , giab008 ( 2021 ). OpenUrl CrossRef PubMed 54. ↵ Bray , N. L. , Pimentel , H. , Melsted , P. & Pachter , L . Near-optimal probabilistic RNA-seq quantification . Nat. Biotechnol . 34 , 525 – 527 ( 2016 ). OpenUrl CrossRef PubMed 55. ↵ Love , M. I. , Huber , W. & Anders , S . Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2 . Genome Biol . 15 , 550 ( 2014 ). OpenUrl CrossRef PubMed 56. ↵ Hao , Y. et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis . Nat. Biotechnol . 42 , 293 – 304 ( 2024 ). OpenUrl CrossRef PubMed 57. ↵ McGinnis , C. S. , Murrow , L. M. & Gartner , Z. J . DoubletFinder: Doublet detection in single-cell RNA sequencing data using artificial nearest neighbors . Cell Syst . 8 , 329 – 337 .e4 ( 2019 ). OpenUrl PubMed 58. ↵ Kowalczyk , M. S. et al. Single-cell RNA-seq reveals changes in cell cycle and differentiation programs upon aging of hematopoietic stem cells . Genome Res . 25 , 1860 – 1872 ( 2015 ). OpenUrl Abstract / FREE Full Text 59. ↵ Marsh , S. , Salmon , M. , Hoffman , P. , kew & Pir , M. S. Samuel-Marsh/ScCustomize: Version 3.2.0 . ( Zenodo , 2025 ). doi: 10.5281/ZENODO.17094431 . OpenUrl CrossRef 60. ↵ Gilbreath , C. , Peng , Y. & Wu , S . Robust detection of gene amplification in formalin-fixed paraffin-embedded samples by fluorescence in situ hybridization . J. Vis. Exp . ( 2024 ) doi: 10.3791/66978 . OpenUrl CrossRef 61. ↵ Tang , Z. , Kang , B. , Li , C. , Chen , T. & Zhang , Z . GEPIA2: an enhanced web server for large-scale expression profiling and interactive analysis . Nucleic Acids Res . 47 , W556 – W560 ( 2019 ). OpenUrl CrossRef PubMed 62. ↵ Heigwer , F. , Kerr , G. & Boutros , M . E-CRISP: fast CRISPR target site identification . Nat. Methods 11 , 122 – 123 ( 2014 ). OpenUrl CrossRef PubMed Web of Science View the discussion thread. Back to top Previous Next Posted November 16, 2025. Download PDF Supplementary Material Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following ecDNA-driven oncogene super-expressors shape immunoevasive tumor microenvironment Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share ecDNA-driven oncogene super-expressors shape immunoevasive tumor microenvironment Kailiang Qiao , Qing-Lin Yang , Tuo Li , Xiongfeng Chen , Zeynep Yazgan , Yoon Jung Kim , Collin Gilbreath , Jun Yi Stanley Lim , Yipeng Xie , Xiaohui Sun , Yang Liu , Yiyue Jia , Zhijian J. Chen , Huocong Huang , Sihan Wu bioRxiv 2025.11.15.688565; doi: https://doi.org/10.1101/2025.11.15.688565 Share This Article: Copy Citation Tools ecDNA-driven oncogene super-expressors shape immunoevasive tumor microenvironment Kailiang Qiao , Qing-Lin Yang , Tuo Li , Xiongfeng Chen , Zeynep Yazgan , Yoon Jung Kim , Collin Gilbreath , Jun Yi Stanley Lim , Yipeng Xie , Xiaohui Sun , Yang Liu , Yiyue Jia , Zhijian J. Chen , Huocong Huang , Sihan Wu bioRxiv 2025.11.15.688565; doi: https://doi.org/10.1101/2025.11.15.688565 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Cancer Biology Subject Areas All Articles Animal Behavior and Cognition (7635) Biochemistry (17697) Bioengineering (13894) Bioinformatics (41951) Biophysics (21455) Cancer Biology (18592) Cell Biology (25507) Clinical Trials (138) Developmental Biology (13380) Ecology (19903) Epidemiology (2067) Evolutionary Biology (24321) Genetics (15610) Genomics (22509) Immunology (17737) Microbiology (40398) Molecular Biology (17182) Neuroscience (88618) Paleontology (667) Pathology (2833) Pharmacology and Toxicology (4825) Physiology (7641) Plant Biology (15158) Scientific Communication and Education (2046) Synthetic Biology (4296) Systems Biology (9825) Zoology (2271)